【摘要】 采用双边遍历扫描(Double-SideTraversalScanning)方法针对图像平面中手指指尖定位问题进行了研究。文中通过在线肤色训练,在HSV色彩空间中对训练区中的彩色图像进行去噪,得到针对当前应用环境下终端用户的肤色统计模型。然后对采集图像进行肤色分割处理,利用得到的二值化图像的零阶矩、一阶矩和二阶矩等参数计算出手部图像重心坐标和方向,使用双边遍历扫描方法对二值化图像进行扫描分析,进而得到了手指指尖坐标位置。不同应用环境下的实验结果表明,这种方法和以往基于长轴长度进行指尖定位的方法相比,不仅具有更高的定位精度,同时对强噪声背景下的抗干扰性也有较大改善。更多还原

【Abstract】 A novel approach based on Double-Side Traversal Scanning (DSTS) is used for fingertip localization for perceptual user interface. In this paper, through on-line skin-color training technique, the filter based on feature distance in HSV color space erases much color noise due to digital vision systems, and a skin-color statistic model of current terminal user is achieved precisely after color information of training region is analyzed. After the captured images are segmented by thresholds achieved from the skin-color model, binary images are obtained. Some parameters of the binary images, its 0th, 1st order moment and its 2nd order moments, are calculated for the centers and orientations of hand region, and then the binary images are scanned using DSTS to achieve fingertip coordinates in image plane. Shown from many experiments’ results, the DSTS can achieve better localization precision of fingertip than former fingertip localization based on the length of long axis, and then, DSTS amends the capability of anti-jamming when the captured color images have the background with hard noise.更多还原